1	Why Current Middleware Fails for Mobile Peer-to-Peer Computing Thomas Kunz Systems and Computer Engineering Carleton University http://kunz-pc.sce.carleton.ca tkunz@sce.carleton.ca
2	Agenda Background Traditional Middleware Requirements for Middleware for Mobile Applications Survey/Evaluation of Mobile Middleware Paradigms Conclusions and Future Work
3	Background We have explored issues related to mobile applications since 1997: WWW transcoding proxy Mobile code for adaptive mobile applications Now: how to incorporate our ideas into middleware for mobile applications What middleware? What extensions? Detailed report available from the WWW, http://www.sce.carleton.ca/wmc/middleware/
4	Middleware Definition: a layer of software that allows peers to interact a cross a network Context: distributed systems Role: coordination and communication (connectivity software) Implements: session and presentation layers Bonus: higher level primitives
5	Traditional Middleware Requirements Network communication Support marshalling and unmarshalling Provide higher level primitives Coordination Components synchronization Components activation and deactivation Reliability Processing individual and multiple requests Support Fault-tolerant and different levels of QoS Scalability Load Balancing (transparent replication, access, and migration) Heterogeneity Integrate elements from various contexts Support interoperability
6	Current Middleware Solutions
7	Current Middleware Solutions (Cont.)
8	Challenges of Mobile P2P Computing Network condition: slow, expensive, disconnected, limited bandwidth, high error rate Mobile devices: limited resources and high diversity Physical mobility: different networks and services
9	An Electronic Shopping System: A Case Study Transparent replication: Application cannot influence the replication decision Location awareness: collecting my order on my way home Dynamic reconfiguration: interacting with servers running different middleware platforms Adaptivity: overloaded server (switching) image/video requested (filtration) Asynchronous: Clients are not blocked during the matching process or receiving notification receipt.
10	Requirements for Next Generation Middleware Dynamic Reconfiguration Detect changes in available resources and reallocate them or notify the application to change its behaviour Adaptivity The ability of a system to recognize unmet needs within its execution context and to adapt itself to meet those needs Context-Awareness Disclose the execution context to the upper layer. The context may include device characteristics, user’s activities, and services Asynchronous Paradigm Decouple the client and server components and deliver multicast messages Lightweight Minimum range of functionality used by most applications
11	Middleware State of the Art Reflective Middleware Reflection approach to support inspection and adaptation Examples: OpenCorba, Open-ORB, and DynamicTAO Tuple Spaces Middleware Tuple space approach to facilitate communication Examples: LIME, TSpaces, and JavaSpaces Context-Aware Middleware Awareness approach to expose the context information Examples: Nexus, Alternis, and SignelSoft Event-based Middleware Event notification approach to support large-scale systems Examples: Hermes, CEA, and JEDI
12	Requirements Addressed by Various Categories
13	Conclusions We have presented current middleware solutions and exhibited their characteristics Why traditional middleware systems fail to support mobile settings We have outlined a set of requirements for next generation middleware We have illustrated the state of the art of middleware for mobile computing and discussed four different classes of middleware We have shown what type of requirements are addressed by each class of middleware
14	Next Steps Better understand mobile application requirements and device/network constraints for specific application domains Select a specific middleware paradigm, experiment with sample system it in our testbed (iPaq/laptop/PCs, IEEE 802.11b wireless LAN and Bluetooth) Suggest and develop appropriate enhancements, based on our prior work
15	Thank you